Amanda Stent PhD '01
Interview from 2018 Multicast Newsletter
An EMT Tackles NLP in NYC
After working at AT&T, Yahoo Labs, and as an Assistant Professor at Stony Brook University, Amanda Stent is currently a product manager at Bloomberg in the CTO Data Science group. She is also on the board for the CRA-W working on the DREU program. She was awarded her PhD from the University of Rochester in 2001, co-authored a book titled The Princess at the Keyboard: Why Girls Should Become Computer Scientists, and volunteers as an EMT.
Tell us a little about yourself- what are you doing now? What were you doing previously?
AS: I’ve had a varied career (so far!). After my PhD at U of R, I did a 6 month postdoc at AT&T Research, then went to Stony Brook University as an assistant professor. After I was awarded tenure, I went on leave back to AT&T Research, which at the time was a “pure research” lab. I had such a great time collaborating with a large team of NLP and speech researchers at AT&T that I gave up my faculty position. After six years, AT&T Research changed its mission and I went to Yahoo Labs, which was being “rebooted” under new Yahoo management. At Yahoo, I became a director of a NLP group, then also of a computer vision/video processing group. After Yahoo closed the Labs, for several months I managed about 12 NLP and recommender systems researchers in an engineering group that supported Yahoo’s news products. My plan was to wait for Yahoo to be sold, then go on the academic job market, but I was recruited to Bloomberg and my husband liked the idea of not moving.
For the past two years, I have been in a product manager role at Bloomberg in the CTO data science group. My primary job is to work with Bloomberg engineering to build core NLP functionality that can be used across the company and by our clients. I also teach NLP within the company to Bloomberg engineers and data scientists, and do some research of my own. Although I was a little nervous to move into this role, it has been very interesting and deeply satisfying so far.
What have your experiences been like in academia vs. industry? Why did you decide to re-enter the industry after being in academia?
AS: Every work environment is unique, and only you can know what matters for you in a work environment. For me, it is: technical innovation should be encouraged; it should be possible to collaborate with smart people on interesting problems; the organization should have a general habit of ethical thinking and practice; and publication should be an option. Also, for me personally, autonomy in my work day is really important. All of these characteristics can be found in academia, in tech, and in other areas that hire PhD computer scientists such as telecommunications, finance, healthcare and the law. Computer science is that relatively rare field where researchers can move back and forth between industry and academia, as long as they keep publishing.
You have authored more than 90 papers on natural language processing and are a co-inventor on more than a dozen patents. Looking back at your career to this date, what contribution in your body of work gives you the most satisfaction?
AS: I loved working on TRIPS as a PhD student! Some of the papers from that work are still very impactful.
I remain deeply interested in grounding (the establishment of mutual understanding between participants in the interaction). One of the highlights of my career so far was working with Susan Brennan and Marie Huffman on a multi-year NSF-funded interdisciplinary grant, looking at grounding and related phenomena in human-human and human-computer interaction.
That said, not all contributions take the form of patents or publications. As computer scientists, software (or hardware) artifacts can be significant contributions. And of course, it is our contributions to the development of others that are often most impactful in the long run. I am enormously happy to have had the privilege of supervising many undergraduate researchers, several MS and PhD students, and numerous summer interns.
What is the next big challenge for NLP?
AS: We are in a heady time for AI, with records being broken left and right. Yet most NLP that is actually used by people amounts to “perception” (speech recognition) or light “NLU” (machine translation). And the most commonly used NLP techniques in industry are still simple ngram counting and maybe some named entity recognition. We are only starting to address the hard problems of machine reading, real human-computer interaction, and information fusion. Tackling these will foster true collaboration between those who work on NLP, representation and reasoning. One of the reasons it’s exciting for me to be at Bloomberg right now is that in the finance industry (where almost every language artifact can be grounded to an event or price) we are uniquely positioned to successfully work on some of these unsolved AI problems.
What inspired you to work on “Humor in Collective Discourse: Unsupervised Funniness Detection in the New Yorker Cartoon Caption Contest?”
AS: Dragomir Radev was a sabbatical visitor in my group at Yahoo Labs. He had previously worked with Bob Mankoff, then long-time cartoon editor at the New Yorker. Mr. Mankoff gave a very memorable talk at Yahoo Labs, and The New Yorker gave us some data and a challenge: to identify automatic ways to prune the list of submissions to the caption contest to help the individual tasked with selecting the finalists. It was a side project, but not unrelated to important problems at Yahoo (such as identifying which comments on a story are worthy of highlighting).
Why does diversity matter in CS?
AS: Computer science is currently one of the biggest drivers of societal change. When only part of society is contributing, then the societal change is lopsided and warped. We have all seen examples of this. We also know that diverse teams are more creative and productive than teams composed of members from a single demographic. So, as a computer scientist who cares about the long-term health of our field, I must care about diversity. For me, this means that part of my service to the research community is outreach. Currently, I am on the board of CRA-W (Sandhya Dwarkadas is also on the board), where I have just switched from co-editing the newsletter to working on the DREU program. I also volunteer as a program evaluator for ABET.
What do you remember fondly about URCS?
AS: The lounge was a great place to eat lunch or take a break, with the inspiration of the bottles from previous PhD students on the wall. I also remember the research group I was in with great fondness: George Ferguson, Mary Swift, Donna Byron, Lucian Galescu, Myrosia Dzikovska, etc. We were a large group (especially for such a small department) and I think we could not have made the dialog systems contributions we did if we had been just working independently.
What do you do in your spare time?
AS: My husband and I both work in New York City but we live in suburban New Jersey with our two cats, Rock and Roll. In my spare time I am a volunteer EMT in my town (it’s great! The opposite of research – you pick a patient up, transport, and drop them off and then it’s completely done! No “future work”!) and a bird watcher.
Knowing what you know now, what advice would you give to current CS students at Rochester?
AS: Graduate school can seem all-consuming – work is life, life is work. But the life of the mind can be balanced. I encourage you to make a list of what matters to you, and then proactively seek opportunities in those areas and say no to everything else. I also encourage you to budget and track your time, and make sure that the “life” buckets aren’t squeezed out by the “research” and “service” buckets.
URCS is a small and collegial program. It’s important to build your professional network through internships, conferences, and volunteering, but also important to enjoy the community of URCS.